Portable FT-NIR spectroscopic sensor for detection of chemical precursors of explosives using advanced prediction algorithms

Near-infrared (NIR) spectroscopy has acquired widespread adoption in various sectors as a result of its benefits over other analytical techniques, the most notable of which is the ability to record spectra for solid samples without any prior manipulation. Furthermore, advances in instrumentation have led to the creation of compact and high-speed spectrometers that can be used in a variety of scenarios, including hazardous materials identification. Fourier Transform NIR (FT-NIR) technology is one of the most useful tools for onsite analysis of chemical and biological substances. Herein, we propose a compact, portable FT-NIR spectroscopic sensor for field measurements, based on commercial broadband light source and spectrometer for detection of chemical precursors of explosives. We mainly focus on four compounds, ammonium nitrate, potassium nitrate, sodium nitrate and urea, some of the best-known chemical precursors of explosives with NIR content. A customized spectral library is constructed, including the forementioned substances under different environmental conditions. We emphasize on two basic factors that can affect the NIR spectra: the relative humidity and the ambient temperature. For the unknown spectrum identification, we evaluate prediction models which involve the use of Random Forest and Support Vector Machine, as well as the Hit Quality Index (HQI) value. The FT-NIR spectroscopic sensor additionally includes an integrated communication module that provides measurement spectra and results to a novel edge computing platform, called DECIoT. We demonstrate the operation of the FT-NIR spectroscopic sensor in real settings under humidity, straight sunlight, and temperature fluctuations, achieving maximum accuracy of 0.96.